ARTICLE

June 6, 2006
SECTION: Section A; Column 1; Editorial Desk; Pg. 21
Witnessing Guilt, Ignoring Innocence?
By David Feige

The police lineup -- in which the anxious eyewitness casts an accusing gaze on a string of sullen men (or women) on the other side of one-way glass -- is as much a staple of actual law enforcement as it is of ''C.S.I.'' and ''Law & Order.'' It is also highly flawed. Yet because of a poorly designed study of lineups in Illinois, much-needed improvements to the process may not be forthcoming.

For some time now, scientists, criminologists and defense lawyers have advocated ending the tradition of having the witness view potential suspects all at once and going instead to a sequential method, in which the witness views suspects (or photographs of them) one at a time. Better still is the ''double-blind'' sequential test, in which the law-enforcement official administering the lineup does not know who is the suspect and who are ''fillers.''

A pilot project in Minneapolis and a generation of peer-reviewed laboratory experiments have concluded that the sequential method has a smaller error rate than old-fashioned simultaneous lineups. New Jersey now uses sequential lineups, and legislatures or commissions in California, North Carolina, Virginia and Wisconsin have endorsed them.

In April, however, Illinois officials released a study showing the new method was less reliable than the traditional one, based on results in three police districts. According to the study, witnesses using the sequential method picked out an innocent person -- someone brought in as a filler -- 9 percent of the time, compared with a 3 percent rate for simultaneous lineups (both figures are oddly below the typical error rate in such studies of 20 percent or more).

While these results have heartened police officials and prosecutors who have resisted changing to sequential tests, the study was so rife with methodological errors and biased analysis that it crossed the line between science and advocacy.

It's hard to understate the power of eyewitness testimony in criminal cases. In thousands of cases every year, testimony of a single eyewitness, uncorroborated by forensic or any other evidence, is used to sustain serious felony charges, including robbery and murder. For years, district attorneys have resisted calls to reform lineups, arguing that because double-blind sequential protocols had never been tested on ''real'' victims outside the laboratory, they shouldn't be adopted in the field. The Illinois project, properly conducted, would have provided an outstanding opportunity to deal with this concern.

The one thing the Illinois experiment did prove was that adoption of double-blind sequential protocols is fairly simple. It also gave more evidence that sequential lineups result in fewer overall identifications: witnesses made ID's in 53 percent of such lineups, versus 62 percent for old-fashioned simultaneous ones. Whether this is an improvement is open to debate.

But the idea that the sequential lineups were less accurate was based on a false premise: that any time the actual suspect was chosen, the result was ''accurate.'' Doing justice in criminal cases is not a matter of identifying the suspect; it's a matter of identifying the actual perpetrator -- who may or may not be the suspect. And while the police are often confident in their judgments, the constant stream of exonerations based on DNA testing and other evidence has starkly demonstrated the gulf between their confidence and scientific accuracy.

Significantly, the Illinois study had no way to control for actual guilt. Indeed, because many of the identified suspects are still wending their way through the system, scientists cannot even assess whether the criminal justice process determined that they were guilty. Nor, apparently, did the researchers have access to the case files, making it impossible for them to determine whether there was any forensic or corroborating evidence at all to support the validity of any particular identification.

Thus to assert that the instances in which the witness singled out the suspect were in fact correct identifications of a criminal is akin to the Alice in Wonderland notion of ''sentence first -- verdict afterwards.''

Another problem with the Illinois pilot project was that it wasn't a true blind study: it allowed police officers who knew the actual suspect's identity to be in the room with the eyewitnesses in the simultaneous lineups, but not in the sequential ones.

And nobody with experience in the justice system would deny that there are plenty of ways, some subtle and some not, for an officer to telegraph to a witness which person in the lineup is the actual suspect. Consequently, comparing an unblinded simultaneous sample to a blinded sequential one is like comparing apples to dolphins.

And then there is the fact that the supposed superiority of simultaneous lineups was detectable in the study's cases in Chicago but not in the nearby Joliet police district. Could it be that big city officers were pushing witnesses toward identifications in circumstances in which their smaller-town brethren are more circumspect?

As with so many aspects of the Illinois pilot study, there is no way to know the truth. But that will not stop police officials and district attorneys from using its pseudo-scientific results as a basis on which to resist the mounting calls for change.

Prosecutors often tell us defense lawyers that they, too, want the fairest possible trials. But when it comes to identifications, it seems they want precisely what the current flawed lineups give them: suspect identifications.

David Feige, a former public defender in the Bronx, is the author of ''Indefensible: One Lawyer's Journey Into the Inferno of American Justice.''